[guijacquemet]

Summary

SWEEP-PP, part A: the small removals of the obsolete-code sweep of this repository (MorphoFeature docs/restart/designs/obsolete-code-inventory-2026-11-10.md §1 and §12, Stage 2 plan §3.7; owner policy of 10 November 2026, Spec ADR 0002 AR10/AR11). It goes into v0.5.1 and is based on main at 8776349 (PP-C, PP-0, PP-1 and PP-2 merged). Part B (#266, stacked on this branch) holds the pairing-editor rebuild and the removal of the design readers (PP-02). One commit per item.

Refs #262. Only part 1 of #262 (the frozen-clock test hygiene) is here. Part 2, the normalization pin move, is a separate PR after this one.

Removed (by inventory item)

Item	What	Commit
#262 part 1	tests/frozen_clock.py skips a module whose import fails only because ipywidgets or matplotlib (the notebook extra) is missing. The chain of causes is followed, and any other import error is still raised. The write test checks a core module's patch always and the notebook module's only when the extra is installed. Regression tests: tests/test_frozen_clock_extras.py.	8f78f44
PP-07	GroupSummary.plane_count, the deprecated alias of channel_histogram_count (before #104). No caller in the nine repositories.	80c0f0b
PP-08 (A7)	source_fingerprint, removed from inventory, from the manifest re-export and from the top level. Tests use source_signature.	28055c7
PP-06	Conversion record: version 3 only. Version 2 was still read; version 1 had a message of its own. Any other version is now refused with one conversion_record_unreadable message that names its version and the fix: convert the originals again into a new folder. The record-level and entry-level projection key is required.	7068ec7
PP-05 (A16)	Run manifest: nothing is read leniently. Removed: the strict flag of recorded_position_id and of the file-entry reader and writer, the version-1 branch of verify_content in to_dict, and the defaults for keys that version 4 always writes (singleton_roles, channel_names, significant_bits, read_in_blocks, channel_names_trimmed, pixel_sizes, settings.histogram, settings.verify_content). sizes_from_data refuses an absent list. In check_sources: the position_id_not_recorded issue is gone (an entry without a position is now identity_differs), and roles are no longer resolved from the image header when an entry has none (such an entry fits no axes: axis_mapping_mismatch).	10bf92e
PP-10	docs/migration-0.3-to-0.4.md, and the links to it in the README, user guide and CHANGELOG.	86c846e

Already gone on main (checked, nothing to do)

• PP-01: undescribed runs (PP-C).
• PP-04: manifest versions 1 to 3 and version_for. READ_VERSIONS == ("4",) (PP-C).
• PP-10, rest of it (PP-C): test_parity_v0_4_0.py, parity_datasets.py, released_code.py, test_removed_in_v040.py and the CI fetch-depth: 0 step. PP-C also renamed test_schema_v2_equivalence.py to test_histogram_goldens.py (science goldens, A6). This PR does not touch it.

Not decided (reported, not changed)

• run --image-type / run_experiment(image_type=) / the manifest's run-level image_type. The inventory lists this under PP-01. PP-1 (PP-1: Fov.image_type per position, partial-mask helper, pinned brick surface #263) documented it in contract §16 as "a record of the command line's review, not an input of a brick". The type is per unit (Fov.image_type), and every described fixture run records it as null. Removing it changes the CLI, the run API and a manifest key, and it overrides a decision made a day ago. Left for the owner to decide.
• PP-03 (draft.migrate_legacy_design): kept. Spec experiment-v1 §7 and conformance P30 still require it; see SWEEP-PP (B): experiment_design.csv is output only, and the pairing card is rebuilt on the draft (PP-02) #266.
• PP-11 stays as decided (A2): position_id is the file stem.

Evidence

The comparison script ran every valid vendored Spec fixture (v01 to v37) through a described run, with masks too where the fixture declares any: 44 cases, 39 written. The refused ones are v04, v05, v06 and v17 (R_HETEROGENEOUS_LAYOUT_NOT_ENABLED) and v29 (mixed rgb/fluorescence, refused by the test helper). It compared main 8776349 with the head of part B (which contains this PR):

• plane_histograms.csv and experiment_design.csv: SHA-256 of the bytes;
• the run manifest, without its run-specific fields (paths, times, sizes, versions);
• the refusal codes of the run and of the mask step;
• mask_representation.json, without mtimes;
• RunInputs.open(...).fovs(): key, condition, repeat group, technical repeat, image type and channel IDs.

Result: 0 differences. Outputs are byte-identical and the refusal codes are the same.

Tests

python tools/check.py on this head 86c846e (Python 3.13, the dev, notebook and zarr extras): ruff check and format clean; pytest 5149 passed, 4 skipped; code rules match their baseline. Skips, all environment-bound:

• test_cli_log.py:166, test_folder_rules.py:199 and :230: this file system tells names apart by case.
• test_discovery.py:665: running with permissions that ignore file modes.

I also ran the three #262 tests with ipywidgets and matplotlib blocked (sys.modules[...] = None). They pass, and on main they fail.

Downstream impact

Pins hold, so nothing breaks until a pin moves past this PR.

• source_fingerprint: CellPatcher mentions it only in a docstring (src/cellpatcher/statistics.py).
• GroupSummary.plane_count: no callers.
• Conversion records of version 2 and manifests missing keys: refused, naming the fix.

The shipped notebook and its RELEASE_REF are unchanged.

A note for the pin-move PR (#262 part 2): PP-07 changes one test in tests/test_report.py. The alias's deprecation test now checks the pooled count. No PlaneRecord(...) constructor was touched, and run/steps.py:229-231 and run/rows.py:33 are unchanged.

🤖 Generated with Claude Code

https://claude.ai/code/session_01G3qp9AxjpGPw9xfWAUJNqc